用可学习的停止策略实现早期抑郁检测，减少误报且提速
Snugi-AI-v2 @ eRisk 2026 Task 2: Early Depression Detection via a Learned Stopping Policy with Sustained Confidence Gate
- 用MLP模型学习何时停止检测，替代固定阈值
- 在500轮内平均第8轮预警，F1达0.73，误报率显著降低
- 适合需要快速、低误报的临床早期干预场景
我们介绍Snugi-AI-v2在eRisk 2026 Task 2中的提交，该任务旨在从Reddit讨论中进行上下文化的早期抑郁检测。核心贡献是训练一个可学习的MLP停止策略，直接优化ERDE50指标，取代以往所有提交中使用的固定分层阈值策略。结合持续置信度门控机制（需连续N=3轮高置信度才触发警报），系统有效抑制了由短期情绪波动引发的误报，同时保持高召回率。整个流程使用冻结的MentalRoBERTa模型编码每条讨论线程，通过MLP分类器将累积表示映射为抑郁概率，并由学习策略决定检测时机。最佳结果实现F1 = 0.73（Run 1）和F_latency = 0.70（Runs 0和3），中位警报轮次为8/500，总评估耗时1小时26分钟，为所有完整提交团队中最短。我们进行了五轮系统性消融实验，涵盖两种编码器变体、四种停止策略和三种门控值，并报告了GRPO策略训练、BDI-II后过滤、MentalLongformer编码及DeBERTa集成等失败尝试。代码已开源。
原文摘要 · Abstract (English)
We describe the Snugi-AI-v2 submission to eRisk 2026 Task 2, the second edition of contextualized early depression detection from Reddit discussions. Our central contribution is a learned MLP stopping policy trained to directly optimize ERDE50, replacing the fixed and tiered threshold strategies used in all prior eRisk Task 2 submissions. Combined with a sustained confidence gate that commits only after N=3 consecutive rounds of high policy confidence, the system reduces false positives caused by transient emotional posts without sacrificing recall. The pipeline encodes each discussion thread with a frozen MentalRoBERTa model, maps the accumulated representation to a depression probability via an MLP classifier, and delegates the timing decision to the learned policy. Our best run achieves F1 = 0.73 (Run 1) and F_latency = 0.70 (Runs 0 and 3), with a median alert round of 8 out of 500, completing the full evaluation in 1 hour 26 minutes, the fastest among all complete-submission teams. We report a systematic ablation across five runs spanning two encoder variants, four stopping strategies, and three gate values, along with negative results from GRPO policy training, BDI-II post filtering, MentalLongformer encoding, and DeBERTa ensembling. Code: https://github.com/chiuyuwen91/erisk-2026
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。